LINGUIST List 11.653

Wed Mar 22 2000

Disc: Phonemic Analysis

Editor for this issue: Karen Milligan <>


  1. Mailbox, Phonemic Analysis, here Larry Trask's contribution # 11.602
  2. Grover Hudson, Phonemic analysis

Message 1: Phonemic Analysis, here Larry Trask's contribution # 11.602

Date: Wed, 22 Mar 2000 09:41:25 +0100
From: Mailbox <>
Subject: Phonemic Analysis, here Larry Trask's contribution # 11.602

 Wow, that was quite a rebuttal! I appreciate the time and effort
 that must have gone into it. I'm not going to try to rebutt Larry as much
 as I'll try (in the spirit of Mark Hubey's suggestions in contribution #
 11617) to come to some kind of common ground between our positions as
 presented earlier.

 First, it seems to me that both Larry and I conflated two separate issues;
 namely, that of phonemic analysis (PA) and that of teaching linguistics to
 beginning students (via PA?). To make matters worse, I also brought POS
 into the argument (or did both of us?) Concerning the latter (i.e.,
 teaching linguistics), it does not behove me to tell anyone how to do
 we all have our experience and our ideas and ideals. Perhaps more
 important, each instructor (I'm using the term here in the sense of "one
 who instructs") has his own personal style, and thus, what works for
one usually
 does not work for another. My main concern in this regard was that I had
 experienced too many linguistics instructors who seem to think that
 teaching any theory is bad for beginning students", and some of Larry's
 remarks sounded to me very much in this direction, as I tried to point

 BTW, Larry, isn't the fact that "an equation should have a quantity on
 side of the equal sign, and ... those two quantities should ... be equal"
 aspect of the theory of algebra/of equations? The same can, I think, be
 said of "the importance of units" and the fact that they have to be of the
 same kind/size, for that, too, is part of (mathematical?) theory. IOW, we
 very often have to teach basic theory (of a related area) before we can
 into "the real thing" (linguistics, in this case).

 There is another issue worth noting here: In my teaching experience the
 students in intro courses were almost exclusively humanities majors, who
 no real idea about POS, scientific method etc. At best, they had vargue
 notions resembling the Logical Positivist notions of scientific inquiry,
 probably picked up in high school. This is why I, at any rate, have been
 pains to present them with a more adequate view of science, and also to
 them understand that in everyday life we don't proceed all that
 from the way we do in science, in the sense that we hold theories and that
 these theories guide most of our endeavors. (More about this later, in
 connection with POS.)

 Before turning to PA and POS, I want to briefly take up a point Mark Hubey
 made (#11.617); namely, that of inexact use of terms. In my case, it was
 perhaps rather "unstated assumptions", but that amounts to much the same
 thing: confusion and/or misunderstandings. Mainly this is so in two
 Dealing with PA as though it was a unified position, and not clarifying
 I meant by "no data without theory". As to the former, the mistake was
 perhaps a natural one for an American linguist, for most of us know only
 PA: "Neo Bloomfieldianism". And Bloomfield was certainly influenced by
 Pos thinking (via Wundt)! Sapir (and Whorf) were never really considered
 Phonemicists by most of us. So, when I said the PA is inadquate and
 incorrect, I was thinking of the positions that held "the separation of
 levels", "once a
 phoneme, always a phoneme", and all that. Forgetting that, particularly
 European linguists, there are many other kinds of PA. Note, BTW, that I
 not arguing against (all of) PA as such, I was arguing against the
 inadequate view of "how we do science", propagated by Log Pos and accepted
 by some (American) practitioners of PA.

 As for the latter, I was thinking in terms of collecting data in
 experiments. That is, I tried to discredit the classical Log Pos picture
 the scientific method that says that we go out and collect data (actually,
 that we perform experiments in order to collect data), and once we have a
 sufficient or adequate amount of those, we try to construct an
 This certanly is not the way linguists collect their data! On the
 we seem to be overburdened by theoretical preconceptions (usually
 ones) before we even start collecting data. Think of Jesuit grammars of
 Amerindian languages like Quechua, or of Japanese, or your own example of
 ergative lgs. Or of the problems Western linguists have with the notion
 "word" in Sino-Tibetan lgs. I remember Pawley (an Australian linguist)
 presenting a paper about a Papuan language, which, he claimed, has some
 forty-five stems (roots?), which are combined to make up all the concepts
 the language "needs".

 Now, you will probably say that all of these are examples of "data without
 theory", but I can equally well say that they are examples of inadequate
 theories being confronted with data they cannot explain. The point is not
 that there are data we can't explain within a given theory (that's old
 the point, at least the one I was and am trying to make, is that what
 us in collecting these data was a theory, one that later turned out to be
 inadequate, to be sure. And that's how "progress in science" comes about:
 in the realization that a given (currently held) theory is unable to
 certain data and that, therefore, it needs to be extended, revisedd,

Now, please, don't think I'm naive enough to think that there are no
"unexplained phenomena" for currently held theories. There are, and in
all areas of scientific enquiry I know of. But, as we all know, constructing
a theory is not an easy matter!

There are some points where Larry is not quite accurate (fair) in his
critique. Thus, I never advocated that linguists should "ape what we
think the physicist (or anybody else) might be doing, and ... allow
ourselves to be hypnotized by ... magisterial pronouncements of some group of
philosophers..." Nor did I say, think, or even imply that "Newtonian
physics is today dismissed by phycisists as a laughable relic." I didn't
know that
it's still taught in basic courses, though it was certainly so taught when
I was an undergraduate. I know that it very adequately explains many
aspects of the physical universe (rockets to Mars, too, but that's technology, not
science), and it appeals to common sense. And there's an interesting
parallel with PA: it, too, appeals to common sense. Perhaps what we
should learn from the phycists is a little more tolerance for what we
may want to call "partially adequate theories" for use in intro ling
courses. As long as they're not patently and obviously false,
like Neo Bloomfieldianism's "separation of levels", etc.

 This brings us to a major (and my last) point, one that has been raised
 repeatedly by various listers (e.g., Ahmad Lohtfi, Carnie, and, if I
 understand him correctly, Moonhawk), and implied by Larry and me: Many of
 us are none-too-tolerant of any ling theory not embraced by ourselves, and
 many of us are also intolerant of any but the latest theory. Perhaps this
 has to do with our late arrival on the scientific scene and the attendant
 insecurity, and with the many popular misconceptions about our discipline.
 The fact that "grammarians" have been around at least since Panini
does not help, because we don't see them (can't sell them to the
public) as scientists. American grad schools, as we know, proselytise
fearfully and
 fiercely, probably as a consequence of the factors mentioned earlier, but
 this certainly reinforces the problem. Like most others, I have no
 solutions here; I can only exhort (proselytize?).

 The implied point I tried to make concerning the similarity between
 scientific enquiry and every day reasoning was made very well by Mark
 Hubley (#11.617), so I won't go into it again here.

 I hope this has cleared up some of the misconceptions caused by my
earalier contribution.


 Peter Menzel
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 2: Phonemic analysis

Date: Wed, 22 Mar 2000 10:15:54 -0500
From: Grover Hudson <>
Subject: Phonemic analysis

Let me try to squeeze in two points in the perhaps overlong discussion
of 'phonemic analysis', especially in response to Menzel (11.429) and
Trask (Vol-11.602, Fri Mar 17 2000). 

First (contra Menzel): data may not exist apart of SOME theory, but
this is a pretty broad use of the term theory. Surely some theory must
often be assumed at the outset in teaching and learning. Just as (I
presume) it is not necessary to get into metaphysical theory, optical
theory, and theories of the neurology of vision and other sense
perceptions before starting to teach basics of the physical sciences
(we reasonably just assume that the elements of these sciences really
DO EXIST as we and our instruments perceive them), in teaching
phonology we can assume that phones exist. This is a relatively
harmless and helpful assumption. (Not ENTIRELY harmless, in fact.)
Similarly, in the study of European history the existence of e.g.
Napoleon will usually be assumed, even though this is now only very
indirectly known -the evidence is quite convincing. And of course
students of the PRACTICE of history should, eventually, thoroughly
understand the indirect nature of such knowledge and the importance of
this for DOING history. 

Later in the study of phonology, indeed, it will be appropriate to
question the existence of phones, and to examine the limits of such
claim. But maybe not in the introductory course --if probably so, to
some extent. 

The complexity (even impossibility) of detailed phonetics REQUIRES that
phonetic data be abstracted as phonemic quite early in the course
-certainly before morphophonemic data come in (authors of some
textbooks seem unaware of the difference). Students should understand
why and how this is done, so they will understand the considerable
evidence of phonology which the theory makes sensible (e.g. allophonic
vs. phonemic awareness, categorical perception, aspects of second
language learning, writing systems), and, particularly, the principled
sense in which a phonetic symbol can represent something which it
doesn't in phonetic writing: e.g. /n/ may represent a voiceless,
dental, palatal, velar, etc. nasal stop.

Second (pro Trask, I think), and the main point: of course phonemics
IS theory, and very important theory for linguistics: that only
CONTRASTIVE aspects of form are lexical, and that resulting allophonic
rules (or constraints) DO EXIST. (Of course by 'phonemics' we don't
mean what might rather be called 'classical American structuralist
phonemics', with insistence on discovery procedures, biuniqueness, etc.
-reasonably a logical positivist theory, as noted by Wojcik in
Vol-11-630.) Some generative phonologists don't accept phonemic theory
in even this basic sense, yet many of these continue to teach
phonemics. Others, it seems, don't teach phonemics -surely with the
result that their students are thoroughly confused and may even turn to
the study of syntax for consolation.

In fact, I don't FULLY accept the theory, because maybe there are
acquisitional/transitional stages of cognition in which what is
contrastive/primary and what is redundant/secondary are not fully
sorted. But the theory that contrast (itself not a simple notion, of
course) is the basis of lexical form (another notion needing
discussion, but not here -the problem is lexical form underlying
alternation) is VERY well supported. Surely it is appropriate to accept
phonemics in this sense as part of the groundwork of linguistics. Most
introductory linguistics and introductory phonology textbooks show that
their authors do so.

Grover Hudson

Department of Linguistics & Germanic

 Slavic, Asian & African Languages

A615 WH, Michigan State University, 

East Lansing, MI 48824-1027

phone 517-355-8471, fax 517-432-2736
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue